From patchwork Thu Aug 3 12:51:36 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Vincent Donnefort X-Patchwork-Id: 13339959 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 62784EB64DD for ; Thu, 3 Aug 2023 12:51:56 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S231768AbjHCMvz (ORCPT ); Thu, 3 Aug 2023 08:51:55 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:47646 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232519AbjHCMvy (ORCPT ); Thu, 3 Aug 2023 08:51:54 -0400 Received: from mail-yw1-x1149.google.com (mail-yw1-x1149.google.com [IPv6:2607:f8b0:4864:20::1149]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id BD7BD359E for ; Thu, 3 Aug 2023 05:51:51 -0700 (PDT) Received: by mail-yw1-x1149.google.com with SMTP id 00721157ae682-583da2ac09fso9730127b3.1 for ; Thu, 03 Aug 2023 05:51:51 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20221208; t=1691067111; x=1691671911; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:from:to:cc:subject:date:message-id:reply-to; bh=4MfGjcR1EbDev2T2p8ePs7pwG/cgdqcucLngAn3GtxI=; b=34BLKcNrv7ILUnQzR0No4vJffC0g7KFnL7oPUm8Qr4pHqROE/NoCWvXyIq9V72epKr k2IfowxMOsUcp0csaey+TohA4rGVPNtQ63ZZdqaLYi9uPEWH9n+N/5k7mNpz6cRRfwli 3cpkrd5G8Ha5mTKMFHsU1OU/L7at+/M5GSgC2wBjJDSCA0BbWENE7C9OdrL3oNALiT7V sWIMCkqHBe8qaA4yzy96Vo3sOFCwKq8anKvMIxOHaUTduMWIk1tCM4wW04zYo+dJu3eQ HskZ3FA5TQEhSES4/+nnIIJJXcg0SuU52R/HwkO6RKaLo/ajZDLdRsnwpE0TB3GRC9hW lzbA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20221208; t=1691067111; x=1691671911; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:x-gm-message-state:from:to:cc:subject:date:message-id:reply-to; bh=4MfGjcR1EbDev2T2p8ePs7pwG/cgdqcucLngAn3GtxI=; b=YtNF0YBacG/ixfwI4D1JGTepb8vGmYel/fHxRbNthk9UOloMTdgQ77QifA0jdEdcyl 9XErQb3hlZ5Uf7aaGkH3rWj5g9CHgHC849yyZoC/T09hTjA5agm17f2Z2uijAfI67tAR 9wpfIiBgrZMDuEJ2Dc5MPfVwqkmRuXwH89UZvjPHIzVQmTrnGAI/SYLPIYIYYLqxQsc2 Fpsjim6bjrc8Rx4WzSMrk6/Nnh7S8g+gB8O954NNrc4cTLUPTKcZYtM2zZv108jrEaGg bPtS5yTq4kMZTvEOktsZhptAsAF+gsbm/G7WSkag4l/ui0lh9eBH5Rd59A6d09gtXqHp JP+w== X-Gm-Message-State: ABy/qLaRn3dwqfNnSzT1L5AaMkZ9MXX36JCwyIKeJ/K8Hi5llCRm7bNS Gb1UQpSakJ85GksOLNFrsNKSPuTqqaz8/3Qi X-Google-Smtp-Source: APBJJlH7EXtgCF49p0I4xP8By4/CaC9Dyx+90B9P+vdpLOLSl9Ghafoz+9odPb1nBO7TShy5Ik2bYeRBhFMwJ5Mr X-Received: from vdonnefort.c.googlers.com ([fda3:e722:ac3:cc00:28:9cb1:c0a8:2eea]) (user=vdonnefort job=sendgmr) by 2002:a81:b615:0:b0:565:a33a:a49f with SMTP id u21-20020a81b615000000b00565a33aa49fmr168360ywh.6.1691067111050; Thu, 03 Aug 2023 05:51:51 -0700 (PDT) Date: Thu, 3 Aug 2023 13:51:36 +0100 In-Reply-To: <20230803125137.1779474-1-vdonnefort@google.com> Mime-Version: 1.0 References: <20230803125137.1779474-1-vdonnefort@google.com> X-Mailer: git-send-email 2.41.0.585.gd2178a4bd4-goog Message-ID: <20230803125137.1779474-2-vdonnefort@google.com> Subject: [PATCH v6 1/2] ring-buffer: Introducing ring-buffer mapping functions From: Vincent Donnefort To: rostedt@goodmis.org, mhiramat@kernel.org, linux-kernel@vger.kernel.org, linux-trace-kernel@vger.kernel.org Cc: kernel-team@android.com, Vincent Donnefort Precedence: bulk List-ID: X-Mailing-List: linux-trace-kernel@vger.kernel.org In preparation for allowing the user-space to map a ring-buffer, add a set of mapping functions: ring_buffer_{map,unmap}() ring_buffer_map_fault() And controls on the ring-buffer: ring_buffer_map_get_reader_page() /* swap reader and head */ Mapping the ring-buffer also involves: A unique ID for each page of the ring-buffer, as currently the pages are only identified through their in-kernel VA. A meta-page, where are stored statistics about the ring-buffer and a page IDs list, ordered. A field gives what page is the reader one and one to gives where the ring-buffer starts in the list of data pages. The linear mapping exposes the meta-page, and each page of the ring-buffer, ordered following their unique ID, assigned during the first mapping. Once mapped, no page can get in or out of the ring-buffer: the buffer size will remain unmodified and the splice enabling functions will in reality simply memcpy the data instead of swapping pages. Signed-off-by: Vincent Donnefort diff --git a/include/linux/ring_buffer.h b/include/linux/ring_buffer.h index 782e14f62201..980abfbd92ed 100644 --- a/include/linux/ring_buffer.h +++ b/include/linux/ring_buffer.h @@ -6,6 +6,8 @@ #include #include +#include + struct trace_buffer; struct ring_buffer_iter; @@ -211,4 +213,9 @@ int trace_rb_cpu_prepare(unsigned int cpu, struct hlist_node *node); #define trace_rb_cpu_prepare NULL #endif +int ring_buffer_map(struct trace_buffer *buffer, int cpu); +int ring_buffer_unmap(struct trace_buffer *buffer, int cpu); +struct page *ring_buffer_map_fault(struct trace_buffer *buffer, int cpu, + unsigned long pgoff); +int ring_buffer_map_get_reader_page(struct trace_buffer *buffer, int cpu); #endif /* _LINUX_RING_BUFFER_H */ diff --git a/include/uapi/linux/trace_mmap.h b/include/uapi/linux/trace_mmap.h new file mode 100644 index 000000000000..41759b9ad9f2 --- /dev/null +++ b/include/uapi/linux/trace_mmap.h @@ -0,0 +1,29 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +#ifndef _UAPI_TRACE_MMAP_H_ +#define _UAPI_TRACE_MMAP_H_ + +#include + +struct trace_buffer_read_page { + unsigned long lost_events; /* Events lost at the time of the reader swap */ + __u32 id; /* Reader page ID from 0 to nr_data_pages - 1 */ + __u32 read; /* Number of bytes read on the reader page */ +}; + +struct trace_buffer_meta { + unsigned long entries; + unsigned long overrun; + unsigned long read; + + unsigned long pages_touched; + unsigned long pages_lost; + unsigned long pages_read; + + struct trace_buffer_read_page reader_page; + + __u32 meta_page_size; /* Size of the meta-page */ + __u32 meta_struct_len; /* Len of this struct */ + __u32 nr_data_pages; /* Number of pages in the ring-buffer */ +}; + +#endif /* _UAPI_TRACE_MMAP_H_ */ diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 52dea5dd5362..cf9a96883a33 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -330,6 +330,7 @@ struct buffer_page { local_t write; /* index for next write */ unsigned read; /* index for next read */ local_t entries; /* entries on this page */ + u32 id; /* ID for external mapping */ unsigned long real_end; /* real end of data */ struct buffer_data_page *page; /* Actual data page */ }; @@ -525,6 +526,12 @@ struct ring_buffer_per_cpu { u64 read_stamp; /* pages removed since last reset */ unsigned long pages_removed; + + int mapped; + struct mutex mapping_lock; + unsigned long *page_ids; /* ID to addr */ + struct trace_buffer_meta *meta_page; + /* ring buffer pages to update, > 0 to add, < 0 to remove */ long nr_pages_to_update; struct list_head new_pages; /* new pages to add */ @@ -929,6 +936,22 @@ static __always_inline bool full_hit(struct trace_buffer *buffer, int cpu, int f return (dirty * 100) > (full * nr_pages); } +static void rb_update_meta_page(struct ring_buffer_per_cpu *cpu_buffer) +{ + if (unlikely(READ_ONCE(cpu_buffer->mapped))) { + /* Ensure the meta_page is ready */ + smp_rmb(); + WRITE_ONCE(cpu_buffer->meta_page->entries, + local_read(&cpu_buffer->entries)); + WRITE_ONCE(cpu_buffer->meta_page->overrun, + local_read(&cpu_buffer->overrun)); + WRITE_ONCE(cpu_buffer->meta_page->pages_touched, + local_read(&cpu_buffer->pages_touched)); + WRITE_ONCE(cpu_buffer->meta_page->pages_lost, + local_read(&cpu_buffer->pages_lost)); + } +} + /* * rb_wake_up_waiters - wake up tasks waiting for ring buffer input * @@ -938,6 +961,10 @@ static __always_inline bool full_hit(struct trace_buffer *buffer, int cpu, int f static void rb_wake_up_waiters(struct irq_work *work) { struct rb_irq_work *rbwork = container_of(work, struct rb_irq_work, work); + struct ring_buffer_per_cpu *cpu_buffer = + container_of(rbwork, struct ring_buffer_per_cpu, irq_work); + + rb_update_meta_page(cpu_buffer); wake_up_all(&rbwork->waiters); if (rbwork->full_waiters_pending || rbwork->wakeup_full) { @@ -1729,6 +1756,7 @@ rb_allocate_cpu_buffer(struct trace_buffer *buffer, long nr_pages, int cpu) init_irq_work(&cpu_buffer->irq_work.work, rb_wake_up_waiters); init_waitqueue_head(&cpu_buffer->irq_work.waiters); init_waitqueue_head(&cpu_buffer->irq_work.full_waiters); + mutex_init(&cpu_buffer->mapping_lock); bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()), GFP_KERNEL, cpu_to_node(cpu)); @@ -3485,7 +3513,7 @@ static void check_buffer(struct ring_buffer_per_cpu *cpu_buffer, return; /* - * If this interrupted another event, + * If this interrupted another event, */ if (atomic_inc_return(this_cpu_ptr(&checking)) != 1) goto out; @@ -4658,6 +4686,13 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer) cpu_buffer->last_overrun = overwrite; } + if (cpu_buffer->mapped) { + WRITE_ONCE(cpu_buffer->meta_page->reader_page.read, 0); + WRITE_ONCE(cpu_buffer->meta_page->reader_page.id, reader->id); + WRITE_ONCE(cpu_buffer->meta_page->reader_page.lost_events, cpu_buffer->lost_events); + WRITE_ONCE(cpu_buffer->meta_page->pages_read, local_read(&cpu_buffer->pages_read)); + } + goto again; out: @@ -4724,6 +4759,13 @@ static void rb_advance_reader(struct ring_buffer_per_cpu *cpu_buffer) length = rb_event_length(event); cpu_buffer->reader_page->read += length; + + if (cpu_buffer->mapped) { + WRITE_ONCE(cpu_buffer->meta_page->reader_page.read, + cpu_buffer->reader_page->read); + WRITE_ONCE(cpu_buffer->meta_page->read, + cpu_buffer->read); + } } static void rb_advance_iter(struct ring_buffer_iter *iter) @@ -5254,6 +5296,19 @@ static void rb_clear_buffer_page(struct buffer_page *page) page->read = 0; } +static void rb_reset_meta_page(struct ring_buffer_per_cpu *cpu_buffer) +{ + struct trace_buffer_meta *meta = cpu_buffer->meta_page; + + WRITE_ONCE(meta->entries, 0); + WRITE_ONCE(meta->overrun, 0); + WRITE_ONCE(meta->read, cpu_buffer->read); + WRITE_ONCE(meta->pages_touched, 0); + WRITE_ONCE(meta->pages_lost, 0); + WRITE_ONCE(meta->pages_read, local_read(&cpu_buffer->pages_read)); + WRITE_ONCE(meta->reader_page.read, cpu_buffer->reader_page->read); +} + static void rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer) { @@ -5298,6 +5353,9 @@ rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer) cpu_buffer->lost_events = 0; cpu_buffer->last_overrun = 0; + if (cpu_buffer->mapped) + rb_reset_meta_page(cpu_buffer); + rb_head_page_activate(cpu_buffer); cpu_buffer->pages_removed = 0; } @@ -5512,6 +5570,11 @@ int ring_buffer_swap_cpu(struct trace_buffer *buffer_a, cpu_buffer_a = buffer_a->buffers[cpu]; cpu_buffer_b = buffer_b->buffers[cpu]; + if (READ_ONCE(cpu_buffer_a->mapped) || READ_ONCE(cpu_buffer_b->mapped)) { + ret = -EBUSY; + goto out; + } + /* At least make sure the two buffers are somewhat the same */ if (cpu_buffer_a->nr_pages != cpu_buffer_b->nr_pages) goto out; @@ -5754,7 +5817,8 @@ int ring_buffer_read_page(struct trace_buffer *buffer, * Otherwise, we can simply swap the page with the one passed in. */ if (read || (len < (commit - read)) || - cpu_buffer->reader_page == cpu_buffer->commit_page) { + cpu_buffer->reader_page == cpu_buffer->commit_page || + cpu_buffer->mapped) { struct buffer_data_page *rpage = cpu_buffer->reader_page->page; unsigned int rpos = read; unsigned int pos = 0; @@ -5871,6 +5935,264 @@ int ring_buffer_read_page(struct trace_buffer *buffer, } EXPORT_SYMBOL_GPL(ring_buffer_read_page); +static void rb_free_page_ids(struct ring_buffer_per_cpu *cpu_buffer) +{ + int i; + + for (i = 0; i < cpu_buffer->nr_pages + 1; i++) + virt_to_page((void *)cpu_buffer->page_ids[i])->mapping = NULL; + + kfree(cpu_buffer->page_ids); + cpu_buffer->page_ids = NULL; +} + +static int rb_alloc_meta_page(struct ring_buffer_per_cpu *cpu_buffer) +{ + if (cpu_buffer->meta_page) + return 0; + + cpu_buffer->meta_page = page_to_virt(alloc_page(GFP_USER)); + if (!cpu_buffer->meta_page) + return -ENOMEM; + + return 0; +} + +static void rb_free_meta_page(struct ring_buffer_per_cpu *cpu_buffer) +{ + unsigned long addr = (unsigned long)cpu_buffer->meta_page; + + virt_to_page((void *)addr)->mapping = NULL; + free_page(addr); + cpu_buffer->meta_page = NULL; +} + +static void rb_setup_ids_meta_page(struct ring_buffer_per_cpu *cpu_buffer, + unsigned long *page_ids) +{ + struct trace_buffer_meta *meta = cpu_buffer->meta_page; + unsigned int nr_data_pages = cpu_buffer->nr_pages + 1; + struct buffer_page *first_page, *bpage; + int id = 0; + + page_ids[id] = (unsigned long)cpu_buffer->reader_page->page; + cpu_buffer->reader_page->id = id++; + + first_page = bpage = rb_set_head_page(cpu_buffer); + do { + if (id >= nr_data_pages) { + WARN_ON(1); + break; + } + + page_ids[id] = (unsigned long)bpage->page; + bpage->id = id; + + rb_inc_page(&bpage); + id++; + } while (bpage != first_page); + + /* install page ID to kern VA translation */ + cpu_buffer->page_ids = page_ids; + + meta->meta_page_size = PAGE_SIZE; + meta->meta_struct_len = sizeof(*meta); + meta->nr_data_pages = nr_data_pages; + meta->reader_page.id = cpu_buffer->reader_page->id; + rb_reset_meta_page(cpu_buffer); +} + +static inline struct ring_buffer_per_cpu * +rb_get_mapped_buffer(struct trace_buffer *buffer, int cpu) +{ + struct ring_buffer_per_cpu *cpu_buffer; + + if (!cpumask_test_cpu(cpu, buffer->cpumask)) + return ERR_PTR(-EINVAL); + + cpu_buffer = buffer->buffers[cpu]; + + mutex_lock(&cpu_buffer->mapping_lock); + + if (!cpu_buffer->mapped) { + mutex_unlock(&cpu_buffer->mapping_lock); + return ERR_PTR(-ENODEV); + } + + return cpu_buffer; +} + +static inline void rb_put_mapped_buffer(struct ring_buffer_per_cpu *cpu_buffer) +{ + mutex_unlock(&cpu_buffer->mapping_lock); +} + +int ring_buffer_map(struct trace_buffer *buffer, int cpu) +{ + struct ring_buffer_per_cpu *cpu_buffer; + unsigned long flags, *page_ids; + int err = 0; + + if (!cpumask_test_cpu(cpu, buffer->cpumask)) + return -EINVAL; + + cpu_buffer = buffer->buffers[cpu]; + + mutex_lock(&cpu_buffer->mapping_lock); + + if (cpu_buffer->mapped) { + WRITE_ONCE(cpu_buffer->mapped, cpu_buffer->mapped + 1); + goto unlock; + } + + /* prevent another thread from changing buffer sizes */ + mutex_lock(&buffer->mutex); + atomic_inc(&cpu_buffer->resize_disabled); + mutex_unlock(&buffer->mutex); + + err = rb_alloc_meta_page(cpu_buffer); + if (err) { + atomic_dec(&cpu_buffer->resize_disabled); + goto unlock; + } + + /* page_ids include the reader page while nr_pages does not */ + page_ids = kzalloc(sizeof(*page_ids) * (cpu_buffer->nr_pages + 1), + GFP_KERNEL); + if (!page_ids) { + rb_free_meta_page(cpu_buffer); + atomic_dec(&cpu_buffer->resize_disabled); + err = -ENOMEM; + goto unlock; + } + + /* + * Lock all readers to block any page swap until the page IDs are + * assigned. + */ + raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); + + rb_setup_ids_meta_page(cpu_buffer, page_ids); + /* + * Ensure rb_update_meta() will observe the meta-page before + * cpu_buffer->mapped. + */ + smp_wmb(); + WRITE_ONCE(cpu_buffer->mapped, 1); + + /* Init meta_page values unless the writer did it already */ + cmpxchg(&cpu_buffer->meta_page->entries, 0, + local_read(&cpu_buffer->entries)); + cmpxchg(&cpu_buffer->meta_page->overrun, 0, + local_read(&cpu_buffer->overrun)); + cmpxchg(&cpu_buffer->meta_page->pages_touched, 0, + local_read(&cpu_buffer->pages_touched)); + cmpxchg(&cpu_buffer->meta_page->pages_lost, 0, + local_read(&cpu_buffer->pages_lost)); + + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); +unlock: + mutex_unlock(&cpu_buffer->mapping_lock); + + return err; +} + +int ring_buffer_unmap(struct trace_buffer *buffer, int cpu) +{ + struct ring_buffer_per_cpu *cpu_buffer; + int err = 0; + + if (!cpumask_test_cpu(cpu, buffer->cpumask)) + return -EINVAL; + + cpu_buffer = buffer->buffers[cpu]; + + mutex_lock(&cpu_buffer->mapping_lock); + + if (!cpu_buffer->mapped) { + err = -ENODEV; + goto unlock; + } + + WRITE_ONCE(cpu_buffer->mapped, cpu_buffer->mapped - 1); + if (!cpu_buffer->mapped) { + /* Wait the writer and readers to observe !mapped */ + synchronize_rcu(); + + rb_free_page_ids(cpu_buffer); + rb_free_meta_page(cpu_buffer); + atomic_dec(&cpu_buffer->resize_disabled); + } + +unlock: + mutex_unlock(&cpu_buffer->mapping_lock); + + return err; +} + +/* + * +--------------+ + * | meta page | pgoff=0 + * +--------------+ + * | data page1 | page_ids=0 + * +--------------+ + * | data page2 | page_ids=1 + * ... + */ +struct page *ring_buffer_map_fault(struct trace_buffer *buffer, int cpu, + unsigned long pgoff) +{ + struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu]; + + if (!pgoff) + return virt_to_page((void *)cpu_buffer->meta_page); + + pgoff--; + if (pgoff > cpu_buffer->nr_pages) + return NULL; + + return virt_to_page((void *)cpu_buffer->page_ids[pgoff]); +} + +int ring_buffer_map_get_reader_page(struct trace_buffer *buffer, int cpu) +{ + struct ring_buffer_per_cpu *cpu_buffer; + unsigned long reader_size; + unsigned long flags; + + cpu_buffer = rb_get_mapped_buffer(buffer, cpu); + if (IS_ERR(cpu_buffer)) + return (int)PTR_ERR(cpu_buffer); + + raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); +consume: + if (rb_per_cpu_empty(cpu_buffer)) + goto out; + + reader_size = rb_page_size(cpu_buffer->reader_page); + + /* + * There are data to be read on the current reader page, we can + * return to the caller. But before that, we assume the latter will read + * everything. Let's update the kernel reader accordingly. + */ + if (cpu_buffer->reader_page->read < reader_size) { + while (cpu_buffer->reader_page->read < reader_size) + rb_advance_reader(cpu_buffer); + goto out; + } + + if (WARN_ON(!rb_get_reader_page(cpu_buffer))) + goto out; + + goto consume; +out: + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); + rb_put_mapped_buffer(cpu_buffer); + + return 0; +} + /* * We only allocate new buffers, never free them if the CPU goes down. * If we were to free the buffer, then the user would lose any trace that was in From patchwork Thu Aug 3 12:51:37 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Vincent Donnefort X-Patchwork-Id: 13339958 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id C6F7CC04A94 for ; Thu, 3 Aug 2023 12:51:57 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S235608AbjHCMv4 (ORCPT ); Thu, 3 Aug 2023 08:51:56 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:47662 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S235591AbjHCMvz (ORCPT ); Thu, 3 Aug 2023 08:51:55 -0400 Received: from mail-yw1-x114a.google.com (mail-yw1-x114a.google.com [IPv6:2607:f8b0:4864:20::114a]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 4DD373588 for ; Thu, 3 Aug 2023 05:51:54 -0700 (PDT) Received: by mail-yw1-x114a.google.com with SMTP id 00721157ae682-583b256faf5so15426247b3.1 for ; Thu, 03 Aug 2023 05:51:54 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20221208; t=1691067113; x=1691671913; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:from:to:cc:subject:date:message-id:reply-to; bh=J4zJKDo0qVw9nnDi02wK0gUvUvTiEvrCWmDoOMueZ+g=; b=cbLncG2zOcIvXhC07TxZZ2AohsMrE3yRXBJp4hWu0igmxn9PNdu7mG8JDPT0CXS1VS OcnNUUtpAwN7WZWEfdwNijwZsczhrLwPTgQ2GCqz4OdLe/u57lnxD7qBX+wpcmSFj3ek gO2XX8FHNk9tUTP0UJJKmy1/zMKQfOdTnwIhwqhXOYrr8zNbOR+PlqROOrESd9N/ryzR dZSu4ETnKqTgZ2WNQjucuJNZm6uszBq52rNjuoPvBt/Tlp+EQLCFJq40mMDOFkzm6Y8F vPln6x71dyzK3tkwcxFWm9QhjUrjc9bhQYl0wMkSfXyNKl/UEOkcab3mFuc+Xx8/E66i WELA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20221208; t=1691067113; x=1691671913; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:x-gm-message-state:from:to:cc:subject:date:message-id:reply-to; bh=J4zJKDo0qVw9nnDi02wK0gUvUvTiEvrCWmDoOMueZ+g=; b=E8PuWPGg1uYYmYy7u0RBezTUyJdGqSOwM8tNyVpvdVjWoLl8s34Ncx6owohwIf6xvv wG+d8PX/DEppebjRAa4JXXn2GPDdzP8R4WbxT6gXa5hJgQFekcahoerzwH8MMGeqHn/6 0fvJZsHTf1f6lq76AlEiAGats3uzT8mGzWikIoov+lBHQL9gZguRNw/kAu+T5uefdw// nG3bHKtqEzZ9XPYUJeDpf9FqWY9yTRdhsf0cOZ5g2TF9MAVYvR5Z/k3LXYIWyury6W26 S9tHG9vMG+dPb/45+F0EBBwJUGC0Qkx97l3r8TagwArnyLJ6kqPxVx8GUuJrCh+lWVYC WjQA== X-Gm-Message-State: ABy/qLYjG5T2MHay+HVittVYU7uWj4p6ghVte9EVXRi1U2MVkhXOI3jk P9ImEqiWiTD6kExEC0lyf/azEfh5FluNuf7D X-Google-Smtp-Source: APBJJlH1og4uqd3nlr+a1kPHamVKYDX95Gpj/Lvn/1DldH8F4AoSWkXDsIn/BaJlVDRjPz5RE8ibfc80VUgKw6f7 X-Received: from vdonnefort.c.googlers.com ([fda3:e722:ac3:cc00:28:9cb1:c0a8:2eea]) (user=vdonnefort job=sendgmr) by 2002:a81:c509:0:b0:583:9d8d:fb0d with SMTP id k9-20020a81c509000000b005839d8dfb0dmr253224ywi.0.1691067113616; Thu, 03 Aug 2023 05:51:53 -0700 (PDT) Date: Thu, 3 Aug 2023 13:51:37 +0100 In-Reply-To: <20230803125137.1779474-1-vdonnefort@google.com> Mime-Version: 1.0 References: <20230803125137.1779474-1-vdonnefort@google.com> X-Mailer: git-send-email 2.41.0.585.gd2178a4bd4-goog Message-ID: <20230803125137.1779474-3-vdonnefort@google.com> Subject: [PATCH v6 2/2] tracing: Allow user-space mapping of the ring-buffer From: Vincent Donnefort To: rostedt@goodmis.org, mhiramat@kernel.org, linux-kernel@vger.kernel.org, linux-trace-kernel@vger.kernel.org Cc: kernel-team@android.com, Vincent Donnefort Precedence: bulk List-ID: X-Mailing-List: linux-trace-kernel@vger.kernel.org Currently, user-space extracts data from the ring-buffer via splice, which is handy for storage or network sharing. However, due to splice limitations, it is imposible to do real-time analysis without a copy. A solution for that problem is to let the user-space map the ring-buffer directly. The mapping exposed via the per-CPU file trace_pipe_raw. The first page is the meta-page and is followed by each page of the ring-buffer, ordered by their unique page ID. It is therefore easy to translate a page-ID to an offset in the mapping. * Meta-page -- include/uapi/linux/trace_mmap.h for a description * Page ID 0 * Page ID 1 ... The mapper must then do what use to be the kernel job: swap the reader with the head. This is done with a newly introduced ioctl: TRACE_MMAP_IOCTL_GET_READER_PAGE. Signed-off-by: Vincent Donnefort diff --git a/include/uapi/linux/trace_mmap.h b/include/uapi/linux/trace_mmap.h index 41759b9ad9f2..4c915f1d0065 100644 --- a/include/uapi/linux/trace_mmap.h +++ b/include/uapi/linux/trace_mmap.h @@ -26,4 +26,6 @@ struct trace_buffer_meta { __u32 nr_data_pages; /* Number of pages in the ring-buffer */ }; +#define TRACE_MMAP_IOCTL_GET_READER_PAGE _IO('T', 0x1) + #endif /* _UAPI_TRACE_MMAP_H_ */ diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index b8870078ef58..cec6f7599a2c 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -6461,7 +6461,7 @@ static void tracing_set_nop(struct trace_array *tr) { if (tr->current_trace == &nop_trace) return; - + tr->current_trace->enabled--; if (tr->current_trace->reset) @@ -8495,15 +8495,31 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos, return ret; } -/* An ioctl call with cmd 0 to the ring buffer file will wake up all waiters */ static long tracing_buffers_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { struct ftrace_buffer_info *info = file->private_data; struct trace_iterator *iter = &info->iter; + int err; - if (cmd) - return -ENOIOCTLCMD; + if (cmd == TRACE_MMAP_IOCTL_GET_READER_PAGE) { + if (!(file->f_flags & O_NONBLOCK)) { + err = ring_buffer_wait(iter->array_buffer->buffer, + iter->cpu_file, + iter->tr->buffer_percent); + if (err) + return err; + } + + return ring_buffer_map_get_reader_page(iter->array_buffer->buffer, + iter->cpu_file); + } else if (cmd) { + return -ENOTTY; + } + /* + * An ioctl call with cmd 0 to the ring buffer file will wake up all + * waiters + */ mutex_lock(&trace_types_lock); iter->wait_index++; @@ -8516,6 +8532,63 @@ static long tracing_buffers_ioctl(struct file *file, unsigned int cmd, unsigned return 0; } +static vm_fault_t tracing_buffers_mmap_fault(struct vm_fault *vmf) +{ + struct ftrace_buffer_info *info = vmf->vma->vm_file->private_data; + struct trace_iterator *iter = &info->iter; + vm_fault_t ret = VM_FAULT_SIGBUS; + struct page *page; + + page = ring_buffer_map_fault(iter->array_buffer->buffer, iter->cpu_file, + vmf->pgoff); + if (!page) + return ret; + + vmf->page = page; + + get_page(vmf->page); + vmf->page->mapping = vmf->vma->vm_file->f_mapping; + vmf->page->index = vmf->pgoff; + + return 0; +} + +static void tracing_buffers_mmap_close(struct vm_area_struct *vma) +{ + struct ftrace_buffer_info *info = vma->vm_file->private_data; + struct trace_iterator *iter = &info->iter; + + ring_buffer_unmap(iter->array_buffer->buffer, iter->cpu_file); +} + +static void tracing_buffers_mmap_open(struct vm_area_struct *vma) +{ + struct ftrace_buffer_info *info = vma->vm_file->private_data; + struct trace_iterator *iter = &info->iter; + + WARN_ON(ring_buffer_map(iter->array_buffer->buffer, iter->cpu_file)); +} + +static const struct vm_operations_struct tracing_buffers_vmops = { + .open = tracing_buffers_mmap_open, + .close = tracing_buffers_mmap_close, + .fault = tracing_buffers_mmap_fault, +}; + +static int tracing_buffers_mmap(struct file *filp, struct vm_area_struct *vma) +{ + struct ftrace_buffer_info *info = filp->private_data; + struct trace_iterator *iter = &info->iter; + + if (vma->vm_flags & VM_WRITE) + return -EPERM; + + vm_flags_mod(vma, VM_DONTCOPY | VM_DONTDUMP, VM_MAYWRITE); + vma->vm_ops = &tracing_buffers_vmops; + + return ring_buffer_map(iter->array_buffer->buffer, iter->cpu_file); +} + static const struct file_operations tracing_buffers_fops = { .open = tracing_buffers_open, .read = tracing_buffers_read, @@ -8524,6 +8597,7 @@ static const struct file_operations tracing_buffers_fops = { .splice_read = tracing_buffers_splice_read, .unlocked_ioctl = tracing_buffers_ioctl, .llseek = no_llseek, + .mmap = tracing_buffers_mmap, }; static ssize_t